In this episode
Tal Peretz, CEO and Co-Founder of Onfire, argues that generic AI has made go-to-market harder rather than easier: once every team can generate personalized outreach at volume, the bar rises and the only remaining edge is data and context. He walks Greg Kihlström through vertical AI built for companies selling to technical buyers — CTOs, CISOs, and CIOs who evaluate tools in Reddit threads, Slack groups, and Discord servers rather than on vendor websites. The conversation covers the three technical layers required to turn those public conversations into pipeline (finding unindexed communities, classifying genuine buying intent, and resolving pseudonymous users to real people), what the handoff to a seller looks like in practice at a public cybersecurity company, and which leading indicators tell a revenue org the approach is working before closed-won revenue arrives.
Key takeaways
- AI raised the bar rather than lowering it. Once every marketing and sales team can produce mass personalization at scale, output volume stops differentiating anyone — the advantage moves to whoever has the right data and context.
- First-party data only tells you what you already know. Your calls and meetings capture your side of the relationship; the decisive conversations — competitor comparisons, internal success plans, buying-committee members you’ve never met — happen in rooms you’re not in.
- Technical buyers evaluate in public, but not where you’re looking. Developers and IT buyers discuss tools in Reddit, Slack, and Discord communities, some of which aren’t indexed on the public web and have to be located manually.
- Intent classification is its own layer. Onfire’s models — built on public LLMs plus proprietary data — separate messages showing evaluation, pricing questions, or requests for recommendations from generic conversation.
- Entity resolution is the hard technical problem. Because users on these platforms are pseudonymous, the system has to trace public footprints across communities to identify the actual individual behind a message.
- The same signal means different things to different sellers. A cloud-security vendor and a developer-tools vendor care about different contexts, customer sizes, and conversations, so customer context has to be layered on top of the classified signal.
- Roughly 80% of a seller’s time goes to manual work. Account prioritization, research, and prospecting flows are exactly what AI is good at; the goal Tal describes is redirecting that time to customer-facing selling, not cutting headcount.
- Revenue per rep is the top-line metric. The test is whether the same team produces more, with leading indicators — new relevant prospects discovered, prospects engaged, meetings booked, opportunities created — arriving well before closed-won revenue.
- Someone in the go-to-market org has to own AI. Tal’s recommended first step is a dedicated owner under revenue or marketing operations with time to think from first principles, followed by a decision about where the data will live — because AI will only be as good as the data behind it.
Chapters
- 0:00 — Activity or intelligence? Framing the AI go-to-market question
- 1:47 — From CTO to CEO: Tal Peretz’s background
- 2:16 — What “vertical AI” means and where horizontal falls apart
- 4:33 — Why first-party data alone can’t see the deal
- 6:48 — Casual developer question vs. genuine buying intent
- 7:44 — Finding unindexed communities and classifying the signal
- 8:29 — Entity resolution behind anonymous handles
- 9:18 — Why customer context changes what counts as a signal
- 9:54 — The handoff: data lake, CRM, and agent prompting
- 11:59 — Leading indicators before closed-won revenue
- 13:40 — Taking the 80% busy work off the rep
- 15:04 — Where B2B go-to-market goes next
- 16:16 — First step: someone has to own it
- 18:03 — Start small, prove ROI, then scale workflows
- 19:18 — What it takes to stay agile
What “vertical AI” means for go-to-market teams
Vertical, in Tal’s usage, means specializing in a specific category or industry rather than building a general-purpose layer. Onfire started with companies selling to technical buyers — a deliberately narrow entry point, because those buyers behave differently from other enterprise personas. The reasoning is competitive rather than technical: when general AI capability is available to everyone, a horizontal tool gives every team the same lift and therefore no advantage. Specialization is where the remaining margin sits.
Why first-party data can’t see the deal
The limitation Tal describes is structural, not a data-quality problem. First-party systems record what happened in your meetings and your calls — what a prospect chose to tell you. But the substance of a B2B evaluation happens elsewhere: competitor comparisons, internal success planning, and additions to the buying committee that a seller may never learn about. Third-party signal from public communities fills that gap, and combining it with first-party context is what Onfire is built to do.
Separating buying intent from noise
Turning a public forum into a pipeline source requires three distinct steps, and Tal breaks them out explicitly. First, locate the communities where technical evaluation actually happens — some are not indexed on the public web and had to be found manually. Second, classify every message: Onfire’s own models, built on top of public LLMs like Anthropic’s and OpenAI’s plus proprietary data, judge whether a message signals evaluation, a request for recommendations, or a pricing question, as opposed to ordinary conversation. Third, identify who wrote it, which on platforms like Reddit means resolving a pseudonymous account to a real person through public footprints left across communities.
What the handoff to a seller actually looks like
Tal’s example is a public cybersecurity company running a data-first strategy: every external data source, Onfire included, lands in their Snowflake data lake alongside Salesforce records and call-recording data, so one store holds the full customer context. Salesforce surfaces the data in the seller’s normal workflow, and the company also connected the store to an AI agent so account executives can prompt against Onfire data directly and build their own workflows. The point is that the signal reaches the rep inside tools they already use, with the agent supplying the interface rather than the intelligence.
Measuring it before revenue lands
Closed-won revenue and expansion are the ultimate KPIs, but they arrive too late to steer. Tal’s leading indicators run in sequence down the funnel: how many new relevant prospects were discovered, how many engaged, how many meetings were booked, and how many opportunities were created. Above those sits a single top-line measure — revenue per rep — which is the number that tests the actual claim: that the same headcount produces more.
Starting without a transformation program
Tal is direct about the failure mode: revenue leaders declare that everything will be AI-native, and then nothing is. What works instead is narrower. Put one person inside the go-to-market organization — revenue operations or marketing operations — in charge of it, someone with time outside the usual workflows to think from first principles. Equip them with a decision about where data will land, since AI is only as good as the data feeding it. Then take one workflow at a time, show the ROI internally, and let adoption compound from proof rather than mandate.
FAQ
What is vertical AI in a go-to-market context? Vertical AI means an AI system specialized to a specific industry or buyer category rather than a general-purpose tool. Onfire’s vertical is companies selling to technical buyers such as CTOs, CISOs, and CIOs, where the relevant data sources and buying behaviors differ sharply from other enterprise segments.
Why isn’t first-party data enough for B2B revenue teams? First-party data captures only what a prospect told you directly. Competitor evaluations, internal success plans, and the wider buying committee are discussed outside your calls — so a first-party-only view misses most of the actual decision process.
Where do technical buyers actually research tools? In public developer communities — Reddit, Slack, and Discord among them. Tal notes that some of these communities are not indexed on the public web and had to be found manually rather than discovered through search.
How do you tell a casual developer question from real buying intent? Through a classification layer. Onfire’s models, built on public LLMs plus proprietary data, distinguish messages that show evaluation of tools, requests for recommendations, or pricing and experience questions from generic conversation, then resolve the pseudonymous author to a real individual using public footprints.
What metrics show this approach is working? Leading indicators in funnel order: new relevant prospects discovered, prospects engaged, meetings booked, and opportunities created — with revenue per rep as the top-line measure of whether the same team is producing more.
Does this replace sales headcount? Tal frames it as empowering the existing workforce. Around 80% of a seller’s time goes to manual work like account prioritization, research, and prospecting; the aim is to shift that time into customer-facing, revenue-generating activity rather than reduce the team.
About Tal Peretz
Tal Peretz is the CEO and Co-Founder of Onfire, a contextual AI platform that helps technology companies decode real-time market signals and drive revenue growth using developer and IT buyer intent data. Prior to Onfire, Tal was Chief Technology Officer at OwnID, a Tel Aviv based startup focused on identity security solutions. Tal served at the 8200 intelligence force and built and scaled advanced data and AI systems focused on large-scale entity resolution and signal intelligence. Drawing on this background, he recognized a fundamental gap in modern go-to-market technology: while AI promised precision, revenue teams were still operating on incomplete, noisy, and outdated data. Under his leadership, Onfire was designed from the ground up to solve the data layer for IT sales by combining structured third-party intelligence with first-party customer context into a continuously updated market map. Tal is married to Shai and is a proud father of two, Dan and Ran.
Tal Peretz on LinkedIn: https://www.linkedin.com/in/tal-peretz/
———- Resources ———-
Onfire:
The Agile Brand podcast is brought to you by TEKsystems. Learn more here: https://aglbrnd.co/r/2868abd8085a9703
We’re proud to be a media partner for #MAICON26 – Oct. 13-15! Learn how AI can power your marketing and business and help you grow smarter. Use code AGILE150 to save! https://aglbrnd.co/r/7fe458ced0f04658
Reach your customers with Reddit. Spend $500 in ad spend, get $500 back in ad credit! Learn more: https://advertalize.com/r/491818c79fb1873f
Chaser is the only Slack-native project management platform that helps teams turn messages into tracked tasks, automate follow-ups, and maintain team-wide visibility, without adopting another tool. Now integrated with Claude and other GenAI tools. Learn more at trychaser.com and use code AGILEBRAND for a 3-month free trial (normal trial is 14 days).
The most influential minds in software, AI, and engineering leadership will be at WeAreDevelopers World Congress North America, September 23-25 in San Jose. Learn more: https://aglbrnd.co/r/60a7299222a7bcf1
Start building your own apps with Replit and get $20 off. Learn more: https://aglbrnd.co/r/93531742a7625a20
Enjoyed the show? Tell us more at and give us a rating so others can find the show at: https://aglbrnd.co/r/faaed112fc9887f3
Connect with Greg on LinkedIn: https://www.linkedin.com/in/gregkihlstrom
Don’t miss a thing: get the latest episodes, sign up for our newsletter and more: https://aglbrnd.co/r/35ded3ccfb6716ba
Check out The Agile Brand Guide website with articles, insights, and Martechipedia, the wiki for marketing technology: https://www.agilebrandguide.com
The Agile Brand is produced by Missing Link—a Latina-owned strategy-driven, creatively fueled production co-op. From ideation to creation, they craft human connections through intelligent, engaging and informative content. https://www.missinglink.company
Transcript
[00:00:00] Greg Kihlstrom: In the push to integrate AI into go-to-market strategies, are we optimizing for activity or for intelligence? Agility requires not just adopting new technologies, but critically evaluating how they deliver contextual intelligence. It’s about adapting your strategy based on smarter signals, not just automating existing processes at a higher volume. Today we’re going to talk about moving beyond the hype of generic AI in sales and marketing to focus on what actually drives revenue. We’re gonna explore the concept of vertical AI, specifically for go-to-market teams selling to technical audiences, and how decoding intense signals from public developer conversations can create a significant competitive advantage. To help me discuss this topic, I’d like to welcome Tal Peretz, CEO and co-founder
[00:01:31] Greg Kihlstrom: at Onfire. Tal, welcome to the show.
[00:01:34] Tal Peretz: Thank you so much, Greg. Thank you for having me. Such a pleasure.
[00:01:37] Greg Kihlstrom: Yeah, looking forward to talking about this. Definitely a timely topic here, so let’s, uh, before we dive in though, why don’t you give a little background on your, on yourself and your role at Onfire?
[00:01:47] Tal Peretz: Yeah, for sure. So, uh, I’m Tal, I’m the co-founder and CEO of Onfire. I’ve been dealing with, I wanna say, AI even if before it was a trend for the last decade, building, uh, sophisticated data models in few areas. Previous to my role in Onfire, I’ve been on the other side as a CTO, leading engineering and product teams. So I understand what is to be on the buyer side and now moving into the seller side in Onfire.
[00:02:16] Greg Kihlstrom: Hmm. Yeah. Great, great. Well, yeah, let’s dive in here. And wanna start with, you know, I, I teed up the, the, the term vertical AI in the, in the intro, and wanna maybe start there. And so you, you draw a distinction between generic, more horizontal AI, and a vertical approach for go-to-market teams. First maybe define what, what do we mean by vertical approach, and then, you know, how do you see w- w- why, why vertical? You know, what, what kinda falls apart in, with the horizontal approach?
[00:02:47] Tal Peretz: Yeah, definitely. So just for the audience to, to be on the same page, so when we say vertical it means that we specialize in specific category, specific industry. In our case, we started with helping companies that sell to technical buyers. So as you can imagine, those, uh, sellers that sell to the CTO, CISOs, CIOs, probably one of the toughest buyer out there. Uh, those technical people, they’re, they have their own places when, you know, they, they try to learn and, and take decision on tools. Like places like Reddit, Slack, Discord, et cetera. And when we started the company we understood that something is shifted.
[00:03:33] Tal Peretz: If we’re talking about AI, you know, everyone knows what it is, everyone understand how they can leverage that. But what m- what we had in mind as a team is basically AI right now make it even harder for marketing and sales team to drive their, uh, let’s say mission and bring more pipeline because everyone can right now do what we call like massive personalization or scale marketing with AI. Um, and right now the, the, the bar became higher. So the only approach that we believe in right now is that if you will go vertical and have the right data and contacts,
[00:04:20] Tal Peretz: this is where the true power will, will, will bring from the AI solution right now. So this is the reason why we started verticalized, and we can jump into it in more depth if you would like, but that’s the high level.
[00:04:33] Greg Kihlstrom: So yeah, and as you mentioned, you know, the, the idea of monitoring public channels like Reddit, Stack Overflow, others isn’t new. You know, that’s, that’s been done, but applying a vertical AI layer is something that’s, that’s new. So, uh, you know, maybe talk a little bit more about, uh, you know, what, what strategic advantage does this data layer first approach give a revenue org over one relying on its own first-party data?
[00:04:59] Tal Peretz: So, so I think, you know, the, the easiest way to describe it is with first, fir- first-party data you only know what you know. You have your own calls, you have your own communications with the customer, and what he told you in the last meeting. But as we all understand sales, the m- the interesting stuff happen where you’re not in the room.
[00:05:23] Tal Peretz: Where they are talking about different competitors, they are looking into some success plan that they have in- internally that maybe they bringing in another, you know, uh, part of the buying committee that you wasn’t been even aware of.
[00:05:36] Greg Kihlstrom: Yeah. Yeah.
[00:06:11] Tal Peretz: And we take this approach into the vertical AI solution, as we described, so they have the best context, both from the third party and the first party. And this is why, you know, we see our, um, our customers and, and brands that we are working with, they have a much better understanding what is happening, and their marketing activities become much more relevant with less of a headcount and a budget, but with much higher return of investment. Um, so I think this is the holy grail right now with what is happening with AI, and we’re trying to do it.
[00:06:48] Greg Kihlstrom: Yeah. So let’s, let’s dive a little deeper then into, uh, into the mechanics of this. You know, how does, uh, there’s lots of conversations on, on, on these platforms as, as anyone who’s ever been to one [laughs] knows. Um-
[00:07:01] Greg Kihlstrom: … how do you differentiate between a developer asking a, a casual question and one that is signa- you know, signaling genuine buying intent?
[00:07:11] Tal Peretz: Yeah. So that, that’s the, I think, like, the one of the m- the, the most complex, you know, uh, problems right now, and maybe let’s break it down to the audience to, to few part. The first part is to understand, first of all, that where those, you know, conversations are happening. So we have been investing a lot of finding those open communities on specific platforms. Some of them are not even indexed on the public web, and we need to drill down manually to find them. So that’s number one.
[00:07:44] Tal Peretz: Once we have that, we basically monitor all of the different messages that’s posted on those specific platforms. And then we have a layer of what we call, like, um, customer classification that are own AI models that, of course, build on top, you know, the, the public LLMs like Anthropic and OpenAI, but also with proprietary data from Onfire. We able to describe if the message is something that’s showing any buying intent, like evaluation of tools, asking for recommendations, asking about pricing or any experience, versus someone that maybe, you know, talking about something generic, like, “How is the weather today?”
[00:08:28] Greg Kihlstrom: Yeah. Yeah.
[00:08:29] Tal Peretz: So once we classify that, and we have the message, the, the last pillar of that is the ability to understand who is the prospect behind the message because, as we all know, those platforms, let’s take Reddit, for example, um, you do not, you don’t use your, uh, real name.
[00:08:50] Greg Kihlstrom: Right. Yeah.
[00:08:50] Tal Peretz: So they’re like, you know, anonymized members.
[00:08:53] Tal Peretz: Um, and, and this is where the true technologies kick in because what we are trying to do here is to find out if there are any public footprint that the users basically left on those public communities, and then based on that, the ability to find out and do the entity resolution of who is the actual individual right now.
[00:09:18] Tal Peretz: So the when– the– when you have those three layers and you add on top of that the context of the customer, because one customer that sells, for example, in our end, the cloud security solution, it’s not the same from a company that sells something for the developers then.
[00:09:38] Tal Peretz: So they care about different contexts. They care about different size of customers. They care about different, uh, I wanna say, conversations. So when you integrated all of that in the context of the customer, this is where the true power reveal.
[00:09:54] Greg Kihlstrom: Yeah. Well, and then, then comes that, that handoff, right? So it’s like, what do you– you’ve, you’ve done all, all of that, which is, you know, there, there’s, as you described, there’s a lot going on there to, to identify that, that buyer. What does the handoff look like then to the, you know, to the sales development rep, account exec, or whoever in practice?
[00:10:15] Tal Peretz: Yeah. S- so today, uh, and, and I can give, like, a practical example that we see from a, a, a public, uh, cybersecurity company.
[00:10:25] Tal Peretz: Um, and, and the way they act on the data is I think like one of the most impressive, uh, you know, workflows that I’ve seen. So they have this, like, um, I wanna say, strategy, where they call, like, uh, data first. So any data that they use, like Onfire, always need to sit in inside their data lake, in their case it’s Snowflake-
[00:10:48] Tal Peretz: … that is connected to the data from Salesforce, the data from their, um, uh, call recording system, and you have one place that holds up the entire data and context of your customers.
[00:11:03] Tal Peretz: And then whenever you go out and looking into Salesforce, you will see the data. If you ask a questions on the specific solutions, you will see the data. But then what they’ve done, they basically hooked it up into Cloud Call Work, and then the account executives are able to prompt and ask questions through Onfire with Cloud.
[00:11:29] Greg Kihlstrom: Ah, nice.
[00:11:29] Tal Peretz: So they have the power of the agent plus the right data and context, so they can build, you know, dedicated workflows, dedicated questions that are super personalized for your, their needs right now.
[00:11:43] Greg Kihlstrom: Yeah.
[00:11:59] Greg Kihlstrom: Let’s talk a little bit about measurement then. You know, how-
[00:12:02] Greg Kihlstrom: … obviously, you know, closed deals and, and stuff like that are gonna be the, the ultimate KPI, but you know-
[00:12:07] Greg Kihlstrom: … what, what leading indicators or other metrics … tells an organization that they’re on the right path with an approach like this?
[00:12:15] Tal Peretz: Yeah. So, so first of all, this is probably one of the things that I, I like the most about the, the category of go-to-market, because everything is measurable.
[00:12:25] Tal Peretz: Uh, you, you touched like the, you know, the, the last, uh, the last mile around, like, the, the actual close one revenue and the expansions. Um, but usually the leading indicators that we see is, you know, first of all, how many new prospects I, I discovered in general-
[00:12:41] Tal Peretz: … that are relevant to, to my specific, um, uh, strategy. That’s one. Second, how many got engaged, and then we see meetings that got booked, opportunities being created, and we have the entire funnel.
[00:12:54] Greg Kihlstrom: Yeah.
[00:13:34] Tal Peretz: … type of examples.
[00:13:35] Greg Kihlstrom: Yeah.
[00:13:37] Tal Peretz: So, so this is how we structure that with our customers today.
[00:13:40] Greg Kihlstrom: Yeah. And so, you know, what, what’s the impact then on the sales team itself? You know, I, I would imagine, is this replacing work that they were doing in other ways? Is this, uh, you know, adding a new dimension? You know, what, what, what-
[00:13:54] Greg Kihlstrom: … does this look like for a day-to-day of a sales team, really?
[00:13:57] Tal Peretz: Yeah. That’s a great question. Um, I could tell you that our standpoint on this is we want to empower the existing workforce.
[00:14:06] Tal Peretz: And we believe that today those sellers, basically 80% of their time is doing what we call, like, manual work or busy work.
[00:14:18] Tal Peretz: You know, prioritize the relevant accounts, do the research on them, have the prospecting flow, a lot of things that AI really excellences. And we want to take out this time from the seller so it can be focused on more revenue-generating activities.
[00:14:36] Tal Peretz: Be customer facing, the ability to close more deals and be present. And, and usually where we’re looking to the metrics, we have a, what you call like a high-level metric or top-line metric around what is the revenue per rep?
[00:14:55] Tal Peretz: And we want to see that we uplift that so the team can do more with the same headcount because of Onfire, because of AI.
[00:15:04] Greg Kihlstrom: Yeah. Yeah. And so then, you know, looking, looking ahead, I mean, is this, is this approach the way that, you know, B2B go-to-market teams are gonna, are gonna do things, uh, you know, in a, in a couple of years? Is this gonna remain very specialized? You know, what, what, what do you see here?
[00:15:23] Tal Peretz: Ha, you know, in, in, in the pace of how things are moving in the AI era, I, I think, uh, I really don’t know. That’s the true answer.
[00:15:32] Greg Kihlstrom: Yeah. Yeah. No, no, totally.
[00:15:59] Greg Kihlstrom: Yeah.
[00:16:08] Greg Kihlstrom: Yeah.
[00:16:16] Greg Kihlstrom: Yeah. Yeah. And, you know, so for those, for those out there listening, um, you know, re- regardless of the, the tooling specifically, you know, what’s your recommendation for marketing and sales leaders to think in more from a t- contextual intelligence standpoint? Uh, you know, what’s, what’s a first step to, to get them started down that path?
[00:16:39] Tal Peretz: Yeah, that’s a great question. I think, like, and, and this is like, you know, um, a few examples that I saw that works pretty good, uh, for modern customers in space is, first of all, you need someone inside a go-to-market organization to own it.
[00:16:53] Tal Peretz: It can be under a revenue operation or marketing operation, but someone that, you know, on a daily basis think about AI, think about optimizations, understand the new tools, and is kind of like out of the, you know, usual workflows, usual days, and you have some time to think about it-
[00:17:13] Tal Peretz: … uh, from, from first principles. So once you have this specific hire and then you equip them with the right strategy around where my data is gonna land, because AI will be as good as your data.
[00:17:29] Greg Kihlstrom: Mm.
[00:17:51] Greg Kihlstrom: Mm.
[00:17:51] Tal Peretz: We s- I, you know, I saw personally a lot of, like, revenue leaders that saying, “Okay, everything is gonna be AI native right now.” And then suddenly nothing is.
[00:18:03] Greg Kihlstrom: Yeah. Yeah.
[00:18:03] Tal Peretz: Because you need to prioritize, and you need to understand, and you need to build enablement, and you need to build maybe some technical solutions to, to support that. Um, so I think the best teams, they’re, they start small. With this like, uh, um, uh, I wanna say dedicated teams that build in inside the organization. They show the value, they show the ROI, they sell it internally, and then you start taking workflow after workflow and build, you know, the specific strategy till you take those 80% off.
[00:18:53] Tal Peretz: I think whether that we were talking about more of how AI is gonna change marketing and performance marketing. I think it’s just in the beginning right now. We are seeing a lot of changes on the sales and email marketing and those kinds of places, but I feel like the next, uh, big, uh, um, change is gonna be in performance marketing, and this is just the beginning around that.
[00:19:18] Greg Kihlstrom: Yeah. Love it. Well, we’ll have to, we’ll have to talk about that in a year then. That sounds great. And, uh, last question for you, uh, what do you do to stay agile in your role, and how do you find a way to do it consistently?
[00:19:30] Tal Peretz: I listen a lot to podcasts like this, and I’m trying to learn on a daily basis. Um, and to see what is like the cutting edge, making sure that I understand it, I play with that. I put time on my calendar to try new stuff and new tools. Even as a CEO that run a lot of functions, I think this is the only way in this, you know, um, what do you call, like AI native era to be relevant.
